Stop of MC/SG package "hangs"?

[Co van Berkel]

Hi,
I configured a one node cluster on a HP9000/800 L2000 server whith HP-UX 11.0.
When I execute the command "cmhaltpkg -v " the state of the package stays "halting"!
Also no response of the "cmhaltpkg" command.
When I look in to the package logfile the followong message is displayed:
########### Node "nlunx3n1": Package halt completed at Fri Jul 2 10:28:
58 METDST 2004 ###########
- CMVIEWCL tells me:
CLUSTER STATUS
axa-uitw-cl up

NODE STATUS STATE
nlunx3n1 up running

PACKAGE STATUS STATE AUTO_RUN NODE
apppkg up running enabled nlunx3n1
mqspkg up running enabled nlunx3n1
orapkg up running enabled nlunx3n1
ntppkg halting halting disabled nlunx3n1

- No error messages in the syslog.log?

Can any body tell me what the problem can be??

Greetings,

Co van Berkel.

[Bharat Katkar · Accepted solution]

Hi Berkel,
Package must be waiting for something to close. Did you see /etc/cmcluster/ntppkg/*.log file. That should let you know where exactly the package state is while halting. May be some of the FileSystem are still in use and unable to unmount it.
If post that log file then it will be more clearer and easier to pinpoint the problem.

Regards,

[G. Vrijhoeven]

Hi Co,

Could you provide the info from the package log file.

If in the customer_defined_halt command section of the controle file a script is started to stop some software and that script hangs, the package state will remain halting. Killing that script can help shutting down the package.

HTH,

Gideon

[Co van Berkel]

Hi,
All the filesystems are umounted and the volume groups are detached.
In the packagelog,
"/etc/cmcluster/ntppkg/ctrl.sh.log" the message "Package halt completed" is displayed!
Now it is 1,5 our later that the command is executed and still cmviewcl says that the package is "halting"!
There are no error messages in the syslog.log.
greetings
Co.

[Dietmar Konermann]

ServiceGuard waits for the package control script to terminate before the status goes 'halted'. Have you already checked if there are still associated processes running?

Try this...
# export UNIX95=
# ps -efxH

Then search for your package control script. That process is typically a child of a cmsrvasstd daemon.

Best regards...
Dietmar.

[Co van Berkel]

Hi,
Try this...
# export UNIX95=
# ps -efxH (This doesn't work???)

I did the following:
- "ps -ef|grep -v grep|grep cmsrv"
root 4749 4723 0 Jun 3 ? 00:02 /usr/lbin/cmsrvassistd


- Than the next:
# ps -ef|grep -v grep|grep -E "cmsrv| 4749"
root 4754 4749 0 Jun 3 ? 00:00 /usr/lbin/cmtaped
root 4749 4723 0 Jun 3 ? 00:02 /usr/lbin/cmsrvassistd
root 4753 4749 0 Jun 3 ? 00:00 /usr/lbin/cmlvmd
root 4752 4749 0 Jun 3 ? 00:01 /usr/lbin/cmlogd
-3 22585 4749 1 10:28:56 ? 00:00
-3 22925 4749 1 10:32:42 ? 00:00
-3 25150 4749 0 11:05:15 ? 00:00

- There i see three processes!!
- I did try to stop three of four package!
What does dis meen?

Greetings
Co.

[Dietmar Konermann]

Oh, looks like you don't have the ps -x enhancement installed.

Nevertheless, you have zombies that are children of cmsrvassistd... that's bad, but obviously the cause for your "halting" paclkage state. Looks like cmsrvassistd did not call wait() for its terminated children.

I would suggest to open a support call with you HP response center for further investigation.

Best regards...
Dietmar.

[Co van Berkel]

Thanks all of you..

Greetings,
Co